Upload
daniel
View
219
Download
5
Embed Size (px)
Citation preview
ORI GIN AL PA PER
Exploring uncertainties in a marketplace for cloudcomputing: a revelatory case study
Sabrina Hauff • Jan Huntgeburth • Daniel Veit
� Springer-Verlag Berlin Heidelberg 2014
Abstract Today, highly standardized information technology (IT) resources such
as storage and processing power are becoming available and affordable to all.
Therefore, IT managers evermore focus on reducing the costs and risks that these
resources entail rather than on the benefits or competitive edge they provide. A
marketplace for cloud computing represents a promising alternative for obtaining
standardized IT resources in a highly flexible and scalable manner. Based on a
revelatory case study of a cloud market project, we show that uncertainties are
prevalent in cloud markets and that the principal-agent theory is an adequate per-
spective to study and explain these uncertainties. By triangulating data from formal
and informal interviews, documentations and project meetings, we develop a second-
level contextual understanding of uncertainties in the relationships between market
operator, cloud providers and users. Our analysis reveals that while cloud market
operators have capabilities to mitigate uncertainties between cloud provider and user,
they also cause new uncertainties from the cloud provider and user perspective. As an
outcome, we present a framework that sheds light on the uncertainty trade-offs
involved in the decision to adopt a cloud market by cloud providers and users.
Keywords Cloud computing � e-Markets � Adoption � Case study �Uncertainty � Principal-agent theory
JEL Classification M15
S. Hauff (&) � J. Huntgeburth � D. Veit
School of Business and Economics, University of Augsburg,
Universitatsstr. 16, 86135 Augsburg, Germany
e-mail: [email protected]
J. Huntgeburth
e-mail: [email protected]
D. Veit
e-mail: [email protected]
123
J Bus Econ
DOI 10.1007/s11573-014-0719-3
1 Introduction and motivation
Cloud computing causes fundamental changes within the IT industry and IT service
delivery. Hence, it has received much attention in research and practice over the last
years. Current trends show that the market for cloud computing has an immense
potential: Gartner expects that the public cloud services market grows from 2012 to
2016 worldwide from $109 billion to $206.6 billion (Gartner 2012). The German
Federal Association for Information Technology, Telecommunications and New
Media (BITKOM) predicts a double-digit growth rate from 2012 to 2016 within the
German cloud computing market, growing from 5.3 billion euro to 17.1 billion euro
in 2016 (BITKOM 2012). Nonetheless, many researchers think that cloud computing
is still in its infancy because challenges like data privacy, security management,
virtual machine migration, or a market-oriented resource management have not yet
received sufficient attention (Armbrust et al. 2010; Buyya et al. 2009b; Zhang et al.
2010).
A research topic that emerged only recently deals with the idea of an electronic
marketplace (e-market) for cloud computing. Buyya et al. (2008) first introduced the
concept of a cloud exchange that allows participants to locate cloud providers and
users with an appropriate offer. Cloud providers can use marketplaces as additional
sales channels to increase the utilization of their existing infrastructure. Cloud users
can benefit from an improved price transparency and reduced search costs for
suitable trading partners since a marketplace can preselect suitable providers based
on a user’s preferences and list them all clearly, to name but a few examples.
However, though there is considerable research on e-market adoption examining the
phenomena in various contexts—most of it focuses on studying e-markets in
scenarios where tangible goods are traded (e.g., Chang and Wong 2010;
Gengatharen and Standing 2005; Joo and Kim 2004; Quaddus and Hofmeyer
2007; Tao et al. 2007).
We believe that the motives for adopting a transparent, two sided and
independently managed marketplace for cloud computing are partly distinct from
those identified in previous research on e-markets. First of all, traditional
economic market theories are not applicable in this context. Especially in
economics, markets have been researched for a long time. They can provide an
optimal allocation mechanism for supply and demand. This idea can be dated back
to Smith (1937) who developed the idea of an ‘‘invisible hand’’ guiding an
efficient resource allocation. Yet, theories about efficient markets only hold under
certain simplifying conditions: full information transparency has to be given and
all information has to be included in the market price, so that no uncertainties are
prevalent (Malkiel and Fama 1970). Beyond these simplifying assumptions,
markets are especially well working for highly standardized goods (Williamson
1985) and it has to be assumed that security and data privacy challenges are
absent. Yet, these conditions are not given in a real-world cloud computing
context, as various information asymmetries and uncertainties that arise from those
asymmetries exist. We think that they can be largely attributed to the interesting
characteristics of a cloud computing product which differentiate it from many
traditional goods traded on marketplaces: As nowadays highly standardized IT
S. Hauff et al.
123
resources have become available and affordable to all, IT managers evermore
focus on the risks and costs that these IT resources entail rather than on the
benefits or competitive edge these computing resources can provide (Carr 2003).
In cloud computing, vendor lock-in and non-availability of a service pose such
risks and thus cause large obstacles in cloud computing adoption, for example.
Besides that, when enterprises use a cloud product, they have to transfer their
business data to the cloud which is under control of the provider. As remotely
hosted data can never be fully protected against unauthorized access, additional
uncertainties like privacy and security concerns can explain the reluctance of some
organizations to engage in cloud computing (Armbrust et al. 2010). In
consequence, the level of involvement and risk regarding cloud computing
products is much higher than for tangible products or highly standardized services
(e.g. electricity or package tours) traded on other marketplaces, where no massive
amounts of data are transferred to third-party hosted infrastructures.
Thus, we propose that the success of a marketplace for cloud computing highly
depends on the robustness of the marketplace’s properties to manage uncertainties
influencing the business relationship between providers and users. This implies
especially that the way the marketplace operator designs, implements and operates
the marketplace is critical for its success. Therefore, we focus on the research
question: How do uncertainties influence the decision to adopt either a transparent,
two sided and independently managed marketplace or a direct provisioning mode
for cloud computing?
Overall, this paper is expected to make the following contributions to the
literature on cloud computing. To our best knowledge, our study presents the first
framework for understanding the uncertainty trade-offs involved in the decision to
adopt a public, two sided and independently managed marketplace for large-scale,
commoditized computing resources such as clouds. This framework extends
previous research on e-markets by examining in-depth the uncertainties surrounding
cloud markets and the way they influence the marketplace adoption decision from a
cloud provider and user perspective. Thus, we provide a comprehensive picture
which takes into account not only all participants, namely providers, users, as well
as the marketplace operator, but also the various relationships between them and the
uncertainties arising out of them. Drawing on principal-agent theory, we show that
agency problems are prevalent in cloud markets and that their management is
critical for the feasibility of a cloud market. From a practical perspective, our
framework can be utilized to inform the design of a marketplace for cloud
computing, e.g. using experience operating a public market from other domains.
The remaining parts of the paper are structured as follows. The next section
positions the paper in the field by introducing the key concepts and the literature-
based pre-understanding of the phenomenon. Section three presents the context of
the study and gives a methodological overview of our revelatory case study. In
section four, our results and interpretations are presented. Finally, the last section
discusses the key findings, theoretical contributions, areas for future research and
practical implications.
Exploring uncertainties in a marketplace for cloud computing
123
2 Theoretical foundation
2.1 Cloud computing: service and deployment model
Cloud computing is a global, distributed, service-centric model for delivering IT-
based solutions which offers to users a new elasticity of resources without paying a
premium for large scale (Bakos 1991). Cloud services provide computing resources
in different levels of abstraction. Infrastructure-as-a-Service (IaaS) provides
hardware resources like processing and storage. Platform-as-a-Service (PaaS)
offers, in addition to the infrastructure, a cloud software development environment
to build, test, and deploy cloud applications (Foster et al. 2008). Lastly, Software-
as-a-Service (SaaS) provides complete applications running on cloud infrastructure
which is completely managed and controlled by the provider. Next to these three
well established service types, the term Everything-as-a-Service (XaaS) is often
used as generalization to show that not only infrastructure, platform, and software is
offered as a service but also Business, Data, or Human, for example (Lenk et al.
2009). The idea of a cloud market—as examined in our study—is to trade highly-
standardized computing resources such as processing capacity and storage.
Therefore, we only focus on IaaS in the following.
Three types of deployment models can be differentiated, namely private, public,
and hybrid clouds which differ in organizational aspects like ownership and the
degree of control while they rely on the same technology (Zhang et al. 2010).
Private clouds are typically owned and managed by a single company. While they
offer the highest degree of control over performance, security, and reliability,
private clouds do not offer the typical advantages of clouds like no up-front capital
costs (Grance and Mell 2011). In contrast, public clouds are typically owned, built
and managed by a third party and available to the general public. While there is no
fundamental difference in the underlying technical concepts compared to private
clouds, the consumer’s control over data, network, and security is limited (Armbrust
et al. 2010). Moreover, hybrid clouds are a composition of the deployment models
described above. The distinct cloud infrastructures remain unique entities, but are
linked to enable data and application portability (Grance and Mell 2011). As the
idea of a marketplace for cloud computing is to match third party cloud providers’
excess capacity and users’ demand of computing resources, within our study we
focus on public clouds.
Armbrust et al. (2010) argue that three aspects are new in public cloud computing
from a user perspective. First, public clouds offer the illusion to users of having
access to an infinite amount of resources when needed. Second, computing
resources can be elastically provisioned and released outward and inward
commensurate with demand in a pay-as-you-go manner. Third, up-front investments
are eliminated allowing companies to start small and increase resources only when
there is an increase in their demand. These aspects promise a faster time to market
(Grance and Mell 2011), lower upfront IT costs (Venters and Whitley 2012) and an
easier scale out (Alford and Morton 2009). Compared to conventional data centers,
public clouds benefit from higher economies of scale due to very large data centers.
Moreover, they exploit higher utilization by multiplexing of workloads from
S. Hauff et al.
123
different organizations, simplify operation and increase utilization via resource
virtualization (Armbrust et al. 2010).
2.2 Marketplaces for cloud computing
A marketplace is a space where goods and services are traded. An e-market is an
inter-organizational information system that is designed to establish provider-user
relationships (Bakos 1998). It facilitates communication and transactions between
its participants, often supported by additional services (Stockdale and Standing
2004). The activities normally are trilateral including users, providers, and the firm
operating the marketplace, often referred to as the intermediary or marketplace
operator (Chang and Wong 2010).
Literature discusses different functions and characteristics of markets (Bakos
1998; Coase 1937; Wigand 1997). Bakos (1998) distinguishes three main functions
of markets; namely matching buyers and sellers, facilitating transactions and
providing an institutional infrastructure. In the setting of cloud economics, these
market functions have to be adapted to the concrete nature of cloud computing. In
concrete, cloud markets aim to match users’ demand of IT resources with supply,
i.e. providers’ excess capacity (Li and Jeng 2010). Second, cloud markets stem from
the facilitation of transactions between the user and provider of IT resources (Buyya
et al. 2008). To fulfill the transaction, the service has to be provisioned to the user
while the payment must be transferred to the provider. Cloud markets typically
incorporate mechanisms for these logistic and settlement activities. To protect users
and providers from the opportunistic behavior of other market participants, cloud
markets normally ensure a certain level of trust, e.g. through including banks issuing
letters of credit, credit reporting bureaus, or rating agencies. Finally the institutional
infrastructure of cloud markets, which provides a legal and regulatory framework to
govern market transactions, is of high importance for the market’s success
(Rossbach and Welz 2011). In particular, this relates to issues about contract law,
dispute resolution, and data privacy protection, but also offers mechanisms for their
enforcement. Seen in light of Wigand (1997), we characterize the cloud market as
competitive, due the variety of buyers and sellers, the homogenous nature of
computing resources, easy entrance to and departure from the market, low switching
costs for consumers and close to perfect information.
Table 1 gives an overview of the literature that deals with marketplaces for cloud
computing. This literature has been gathered by conducting a literature review in the
top eight IS journals based on the Association for Information Systems (AIS) Senior
Scholars’ Basket of Journals (AIS 2011) using the search strings ‘‘market’’ or
‘‘trading platform’’ in combination with ‘‘cloud computing’’ in a full text search. We
then manually checked all papers for their relevance. Since this yielded almost no
results, we conducted a broader literature review using the search engines
EBSCOhost, ScienceDirect, ProQuest, and Google Scholar. Thereby, we used the
search terms ‘‘market’’ or ‘‘trading platform’’, again both in combination with
‘‘cloud computing’’, and, in the case of Google Scholar, also the terms
‘‘marketplace’’ and ‘‘cloud computing’’ as well as ‘‘market exchange’’ and ‘‘cloud
computing’’ to further limit the results. Moreover, we conducted a backward and
Exploring uncertainties in a marketplace for cloud computing
123
forward search with all identified papers. As can be seen, the topic has not received
much attention so far. The focus of the research is on generally motivating the topic,
on architecture and on technological aspects like standardized application
programming interfaces (APIs), software components that handle the reservation
and allocation of resources, and other implementation issues. However, to our best
knowledge no study exist which empirically examines the adoption of a marketplace
for cloud computing. We aim to fill this research gap through our study, focusing on
the role of uncertainties that are prevalent between the different parties of a
marketplace scenario.
There is considerable research on e-market adoption (e.g., Chang and Wong
2010; Quaddus and Hofmeyer 2007; Rask and Kragh 2004). As e-marketplaces
evolved from EDI systems (Angeles 2000) and are supported by information and
communication technologies (Malone et al. 1987), researchers have drawn possible
determinants of e–market adoption from these areas.
We believe that the motives for adopting a marketplace for cloud computing are
partly distinct to those identified in previous research on e-markets. On the one
hand, uncertainties play a major role for a decision-maker whether to adopt cloud
computing or not and thus explain the reluctance of some organizations to engage in
cloud computing (Armbrust et al. 2010). On the other hand, IT is nowadays the
backbone of commerce and its core functions—storage, processing and network—
have become available and affordable to all. As argued by Carr (2003), if resources
are essential to competition but inconsequential to strategy, the uncertainties they
create become more important to companies than the advantages they provide. The
diminishing role of IT has important implications for understanding the adoption of
a marketplace for cloud computing. We propose that the market success of a cloud
market will highly depend on the ability of the market operator to manage and
mitigate uncertainties between cloud providers and users compared to bilateral
cloud service scenarios. Today, disruptive IT innovation can paralyze a company’s
ability to produce its products, deliver its services and connect with its customers
(Carr 2003). Hence, IT managers—at least for infrastructure technology—often
focus on vulnerabilities, rather than opportunities for gaining competitive advan-
tage. In many cases, this creates an atmosphere of fear-driven IT-management rather
Table 1 Literature regarding marketplaces for cloud computing
References Topics
Buyya et al. (2008) Motivation, architecture
Buyya et al. (2009a) Architecture, technologies
Buyya et al. (2009b) Motivation, architecture
Krieger et al. (2010) Motivation, challenges, architecture, technologies, API, prototype
Li and Jeng (2010) Architecture, negotiation mechanisms
Garg et al. (2011) Requirements, architecture, prototype
Breskovic et al. (2012) Standardization, SLAs
Calheiros et al. (2012) Architecture, prototype
Grilo and Jardim-Goncalves (2013) Architecture
S. Hauff et al.
123
than a challenge oriented leadership. However, little is known about cloud-specific
uncertainties derogating the provider-user relationship and cloud-market-specific
uncertainties introduced and mitigated by the marketplace operator. Therefore, we
will focus on these aspects in the following.
2.3 Uncertainties surrounding cloud marketplaces
We draw on the principal-agent theory to develop a legitimate frame and potentially
useful vocabulary to enter the field (cf. Fig. 1). We have chosen principal-agent
theory because it particularly allows us to address the relations between the cloud
provider, user and market operator. It focuses on the existing information
asymmetries in these relations and the arising uncertainties and involved trade-
offs when parties evaluate whether or not to enter into a contract. Other theories are
not so suitable in shedding light on these issues. For example, Eisenhardt (1989a)
contrasts several organizational theories with agency theory, like the contingency
theory or the transaction cost theory. While they share some underlying
assumptions, several others are not met by all these contrasted theory. We will
not discuss all of them in detail here (see Eisenhardt (1989a) for more information).
However, especially the transaction cost theory has similar assumptions like
information asymmetries, goal conflicts, bounded rationality, or self-interests of all
parties. Yet, it differs in the aspect of risk preference by assuming risk neutrality as
opposed to an often assumed risk aversion of the agent. Moreover, it offers an
explanation as to which governance structure (market vs. hierarchy or hybrid form)
is chosen based on the occurring transaction costs to reach an agreement between
parties that exchange goods. Transaction costs can be divided into production costs
and costs associated with operation risks and opportunism risks. In addition, the
transaction cost theory deals with demand uncertainty and market volatility as
sources of uncertainty or economic theories that represent a future benefit as a form
of uncertainty (Williamson 1985). Before we contrast principal-agent theory and
transaction cost theory, we first introduce the former one in detail.
Fig. 1 Pre-understanding of uncertainties surrounding cloud markets
Exploring uncertainties in a marketplace for cloud computing
123
Agency problems befall almost all kinds of economic transactions where bounded
rationality, fears of opportunism and information asymmetry exist (e.g., employment:
Schepers et al. 2012; relationships of psychiatrists with industry: Appelbaum & Gold
2010; financial policy: Mugge 2011). An agency problem occurs when one party (the
principal) delegates work to another party (the agent) and the welfare of the principal is
affected by the choices of the agent. Principal-agent theory applies to any variation of
relationships where one party is offering a service and the other party is providing
compensation for the service. Agency problems can arise from two sources, adverse
selection and moral hazard. Adverse selection arises due to a pre-contractual
misrepresentation of providers’ true abilities and qualities as well as the presentation of
false service quality information (Akerlof 1970). In the context of cloud computing,
providers can be seen as agents and users take the role of principals. Thus, while providers
have hidden information about their true qualities, cloud users cannot easily determine a
suitable cloud provider out of a pool of high-quality and low-quality providers. Thereby,
the quality of a provider can for example relate to the effort he puts into fulfilling agreed
upon service level agreements (SLAs) as well as the quality of his data centers. A moral
hazard occurs due to a post-contractual shirking of the provider by not accomplishing a
task as promised (Jensen and Meckling 1976). Even more, it also includes provider’s
deliberate deception of users since they cannot easily evaluate providers’ performance.
Principal-agent theory suggests three measures to overcome the agency problem.
Agents can design and send signals to the principal to expose private information about
their true qualities (e.g. in-depth information about how they operate their service)
(Bakos 1997). The principal examines these signals and may change concerns on using
the agent’s service (Rao and Ruekert 1994). Incentives are designed to prevent
opportunistic behavior of the agent by making any hidden actions irrational or costly
(Pavlou et al. 2007). They can help to align the interests of cloud users and providers
more closely. For example, the user can offer monetary incentives to the provider based
on the quality of the outcome (Vermillion et al. 2002). Finally, principals can also
reduce uncertainties by monitoring the agents’ activities, e.g. by auditing and
controlling the principal randomly (Strausz 1997). In a nutshell, while signaling
provides one solution to adverse selection problems, incentives as well as monitoring
can especially help to resolve moral hazard problems (Mishra et al. 1998).
Overall, the principal-agent theory can help us to analyze the relationship
between cloud users, cloud providers and the market operator as follows: First of all,
it can guide us to understand the potential uncertainties which may exist in the cloud
provider-user relationship by revealing the motives and interest of users and
providers and by interpreting the information asymmetries between the two parties.
Especially the quality of cloud services can hardly be evaluated by the user at any
point in time since the technological details and the behavior of the marketplace
operator and the cloud provider are hidden from him. Therefore, these services are
best described as credence goods. Cloud users’ on-going assessment whether or not
to continue using a cloud service whose quality is hard to discern puts special
emphasis on the importance of uncertainty in our scenario. Large information
asymmetries characterize the prevalent situation in today’s cloud computing
landscape: While the cloud provider has perfect information about his services, the
user does not know about many aspects like future availability of a service, the way
S. Hauff et al.
123
his data is used by the cloud provider, etc. Even more, the user has no chance in
gaining access to this information, even if he has a high willingness to pay for it
(e.g. for monitoring services). The described situation relates well to the concepts of
hidden information and hidden action of the principal-agent theory and is thus
considered as a well-suited theoretical perspective to guide our data collection and
analysis. Principal-agent theory has been successfully used as a theoretical lens to
study uncertainties with regards to search goods (Pavlou et al. 2007) and experience
goods (Dimoka et al. 2012) traded on online markets. In contrast to cloud services,
the quality of the search and experience goods can be either evaluated before the
purchase decision or upon consumption (Nelson 1970). Therefore, we believe that
principal-agent theory is an appropriate lens to study bilateral information
asymmetries between different actors on the cloud market. Moreover, the
perspective allows us to analyze how the introduction of a marketplace influences
this relationship. Yet, a marketplace does not only reduce uncertainties between
cloud users and providers, it also establishes two new agency problems: one
between the user (principal) and the marketplace operator (agent) and another one
between the provider (principal) and the marketplace operator (agent). As in these
relationships goal conflict, fears of opportunism and asymmetric information are
again prevalent, the principal-agent theory can be also utilized to interpret the
uncertainties introduced by the marketplace operator and to understand how these
uncertainties can be successfully reduced.
The transaction cost theory can be clearly demarcated from this: In this theory,
uncertainties are described as arising from market volatility and demand
uncertainty. Even more, the focus of the two theories is different: While the output
of the transaction cost theory is a decision about which governance structure should
be favored (hierarchy vs. market or hybrid forms), the principal-agent theory offers
insights into the relationship of a principal and an agent and investigates what
tactical situations might occur between them by considering potential conflicts,
trade-offs, incentives and control mechanisms. We are especially interested in the
latter point and thus see the principal-agent theory as very valuable for guiding our
research. Yet, our intention is not to test the principal-agent theory in the context of
cloud markets or find instantiations of the concepts in our data. Rather, the
principal-agent theory can be seen as a ‘‘snapshot of our theoretical sensitivity’’
(Sarker et al. 2012, p. 320) to guide data collection and interpretation.
3 The context of study and methodological overview
3.1 Context: case background
We examine the acceptance of a marketplace for cloud computing in the unique
context of a financial marketplace operator specialized in trading shares and other
securities (MarketMaker) who is located in the EU. MarketMaker has decided to
launch a marketplace for IaaS-resources after a 1-year evaluation phase. In this
evaluation phase, a prototype was set up and cloud providers and users were
involved to examine the feasibility of such a marketplace. We name the prototype
Exploring uncertainties in a marketplace for cloud computing
123
CloudMarket in the following. During this phase, we were able to accompany the
marketplace operator and conduct this case study by in-depth experience at any
stage of the evolution of the project.
CloudMarket is a trading platform for highly standardized computing resources.
MarketMaker provides this platform and thereby has the role of a neutral third-party
marketplace operator. It acts only as an intermediary between cloud providers and
users and therefore does not sell own cloud resources. The marketplace mainly
arouses interest among medium and large sized companies since it allows high
volume trading of cloud resources. In the following, we explain the functionality of
CloudMarket in detail and outline the services MarketMaker offers as intermediary.
To become a player on the marketplace, cloud providers and users have to
fulfill certain admission criteria. Cloud providers have to prove that they offer
high-quality services with respect to data security, privacy and reliability based
on certificates issued by third party auditing. Cloud users have to prove that they
have sufficient liable equity to pay for cloud services. In addition, both cloud
providers and users have to agree to the terms and conditions as well as the code
of conduct established by MarketMaker. Once they have been accepted as
marketplace participants, the cloud providers can enter bids while the cloud users
can enter requests for resources. CloudMarket offers two types of markets: a spot
market, where cloud services are sold and delivery takes place immediately, and
a derivatives market, where a gap between trading and delivery is present. The
spot market allows for a flexible, short-term allocation of resources with a
referenced price while the derivatives market enables securing prices through the
hedging of future demand. Thus, it also offers an indicator of future prices. In
each of these markets, computing power and storage resources can be traded. The
definition of standards for IaaS resources as well as contracts is one key
requirement for the marketplace. It enables transparency and allows for a
comparison of services on the one hand and reduces the effort for negotiations on
the other hand. As currently no standards exist, MarketMaker wants to set
standards for each cloud service class. While technical parameters like I/O-
performance and security requirements are fixed, the participants can choose
between predefined values for tradable units, SLAs, the allocation model, and the
product location. The SLAs are kept very simple and define the quality of service
to be expected, metrics for user support, and a refund policy in case SLAs are
violated. The allocation model can have two types: open and fixed end. Open end
means that the contract period is unlimited, has a certain cancellation period for
the contract partners, and includes a scalable function. It allows users an increase
of their contracted resources up to a predefined size under the same conditions. In
a fixed end allocation model, the contract duration is fixed, no cancellation period
exists, and no scalable function is included. Concerning the product location, the
CloudMarket differentiates between data centers located in the European Union
and data centers located within US territory. All these variable parameters can be
combined. To give an example of one occurrence: A product from the asset class
‘‘storage’’ might have a tradable unit of one terabyte, is located in the EU, the
allocation model is open end, and the SLA used is ‘‘SLA1’’ which is the highest
quality available on the marketplace.
S. Hauff et al.
123
CloudMarket supports four phases: pre-trading, trading, clearing, and settle-
ment. We explain the phases and the services MarketMaker offers in each of
these phases. Aim of the pre-trading phase is to allow for an adjustment of the
product portfolio as well as specific parameters, e.g. the adjustment of product
benchmarks or bandwidth due to technological progress. In the trading phase, a
provider or a user can select such a product and enter the quantity he wants to
trade on the marketplace as well as the price he is willing to pay or sell the
resources for. A two-sided auction mechanism is used to determine the matching
of bids and asks. Once the trading has been done, the trade has to be cleared. In
the spot market, the participants clear bilaterally, e.g. via a clearing bank, while
in the derivatives market, a central counterparty (CCP) handles the clearing.
A CCP is an entity which steps in between user and provider. Thereby, it acts as
provider for each user and as user for each provider. Introducing a CCP has
efficiency related reasons because every participant has only one party to interact
with. Next, the settlement takes place. In this step, the user and the provider who
have entered into a contract are connected dynamically and securely by a self-
developed software. Once the connection is established, the consumer can utilize
the resources. At the same time, the settlement software is also monitoring the
provider to check whether he is fulfilling the contract. In case he is violating the
agreement, the refund policy as defined in the SLAs takes effect. This is an
improvement from the current situation where users have to monitor the service
themselves and make claims against the providers.
While the CloudMarket project team has clear ideas about the overall design and
has developed a prototype with a technology partner, a fully implemented and
launched solution of the marketplace does not yet exist.
3.2 Methodology
We believe that the CloudMarket project is a revelatory, unique and exemplary
source of insight on our topic (Yin 2009). MarketMaker is a successful
marketplace operator in the financial sector with a strong reputation. As
trustworthiness of the marketplace operator may be seen as a necessary condition
to establish a cloud market, the CloudMarket project is seen by both providers and
users as a serious alternative to obtain and dispose computing resources. Thus, the
ecology of the CloudMarket project team and its potential business partners
represents a fascinating context where the topic of interest could be investigated in
depth. The goal of our study is to generalize our empirical insights to a theory on
the phenomenon under investigation based on analytical generalization (Yin
2009).
Our engagement with the CloudMarket project contributed to our broader
understanding of the cloud market business case. The first author was
accompanying the project scientifically for half a year. She was present at team
meetings and had access to all internal documents. In addition to reviewing
internal documents, many informal conversations with the team members helped
her to establish a fundamental knowledge of how the marketplace will work and
to understand the opportunities and risks involved in this project. Even more, she
Exploring uncertainties in a marketplace for cloud computing
123
was allowed to participate in some of the meetings with potential consumers and
providers which provided her with additional valuable insights from the
participants’ perspectives.
Moreover, the first author conducted twelve formal interviews with project team
members, cloud providers and users. Overall, twelve interviews were conducted
between April and July 2012: six interviews on the cloud user side, four interviews
on the cloud provider side, and two interviews with representatives of the
marketplace operator. The companies interviewed belong to different branches and
the interviewees held different positions. Aim of this spread of interview partners
was to gain as many insights as possible and triangulate data from different sources
as suggested by methodologist (Patton 2002).
Table 2 provides an overview of the interviews conducted, describing the
industry of the company interviewed, the position of the interviewee in his
company, whether the interview was carried out face to face or via telephone, and
information about the length of the interview.
Each interview was recorded and conducted based on an interview guide (Yin
2009). The interview guide kept interactions focused while allowing individual
experiences to emerge and thus, best used the limited time available in the interview
situation (Gordon 1980). Interview partners were asked questions about their
understanding of cloud computing, their current usage or provision of cloud
resources as well as the opportunities and risks of using cloud computing from their
perspective. Moreover, the marketplace prototype was introduced and the oppor-
tunities and challenges it offers were discussed in depth. The interview guide served
as a ‘‘reminder regarding the information that needs to be collected’’ (Yin 2009,
p. 86). After each interview, the interview guide was adapted in case new
uncertainty concepts have emerged.
Overall, our approach can be characterized as interpretive in that sense that it
takes the interview partners’ experiences with cloud computing to develop a
‘‘second-order theoretical understanding’’ of uncertainties surrounding cloud
markets (Lee 1991; Sarker et al. 2012). In doing so, we applied a less procedural
version of the grounded theory methodology as proposed by Sarker and Sarker
(2009) to develop our understanding of uncertainties surrounding cloud markets.
This involved constant comparison of the data ‘‘[…] to see if [they] support and
continue to support the emerging categories’’ (Holton 2007, p. 277) and then to link
this evolving set of concepts to higher level categories of uncertainties (Walsham
1995). This process led to the discovery of cloud- and market-specific uncertainties
surrounding the market. Finally, based on the experiences of our interview partners,
we identified capabilities of the market place that could help to reduce the
uncertainties surrounding cloud markets (Walsham 1995). Once, concepts and
patterns emerged we enfolded previous literature to support our interpretations
which involved to ask ‘‘what is this similar to, what does it contradict, and why’’
(Eisenhardt 1989b, p. 544).
S. Hauff et al.
123
4 Interpretation and results
In the following, we present our results and interpretations of uncertainties prevalent
in cloud markets and ways to successfully mitigate these uncertainties. We start
with cloud-specific uncertainties which exist between the cloud provider and user.
Next, we discuss capabilities of the market operator to reduce these uncertainties.
As the way the marketplace platform has been designed, implemented and is now
operated by the market maker not only reduces but also introduces new
uncertainties, we analyze cloud-market specific uncertainties between the market
Table 2 Interview partners
Role in
the
market
Industry Position of interviewee(s) Way of
communication
Length of
interview
(min)
User Information technology Interview partner 1
Enterprise Account Manager
& Partner Relations
Face to face 40
User Bank Interview partner 2
Manager of the Architecture
Department
Interview partner 3
Enterprise Architect
Face to face 60
User Logistics Interview partner 4
Management position in
R&D
Telephone 60
User Automobile Interview partner 5
Leader Vitas & Cloud
Computing,
Interview partner 6
Assistant of Cloud Team
Telephone 40
Provider Internet service provider Interview partner 7
Chief Operating Officer
Telephone 50
Provider Internet service provider Interview partner 8
Product Manager
Telephone 45
Provider Internet service provider Interview partner 9
Management position in
R&D
Telephone 40
Provider IT-consultancy, IT-services Interview partner 10
Business Developer
Telephone 55
Market
operator
Equity-, equity index-,
interest rate derivatives,
clearing services
Interview partner 11
Project leader of the
CloudMarket team
Interview partner 12
Consultant—responsible for
the functional work stream
of the project
Face to face 45
Exploring uncertainties in a marketplace for cloud computing
123
operator and its participants up next. Finally, we discuss how a cloud market
operator can successfully mitigate these cloud-market specific uncertainties. The
outcome is a framework which can be used by cloud market operators, cloud
providers and users to understand and successfully manage uncertainties surround-
ing cloud markets.
4.1 Cloud-specific uncertainties
Cloud-specific uncertainties arise through user’s disbelief about the cloud provider’s
ability (adverse selection) or willingness (moral hazard) to provide the computing
resources as agreed upon (Pavlou et al. 2007). Interviewees expressed several types
of concerns that may hinder the adoption of cloud computing. An overview of the
identified uncertainties is depicted in Fig. 2. The concerns presented in the
following are not specific to cloud markets but may occur in all kinds of cloud
computing service scenarios.
First of all, interviewees argued that users have information privacy concerns in
adopting cloud computing. According to Westin (1967, p. 7), privacy refers to ‘‘the
claim of individuals, groups or institutions to determine for themselves when, how
and to what extent information about themselves is communicated to others’’. In the
context of cloud computing, users want several types of data to be protected by the
cloud provider. These include identity information, transaction histories, and,
probably most important, all business critical data which might be transferred to the
provider for processing or storage (Takabi et al. 2010).
‘‘[…] we have repeatedly tested cloud usage in certain business areas.
However, we have repeatedly shied away because we realized that some
internal bank data would be stored in the cloud. We did not want to take this
risk.’’ (interview partner 3)
Privacy concerns arise in both types of agency problems: Ex ante, users may find
it difficult to analyze a cloud provider’s mechanisms to ensure data privacy. Ex post,
a user cannot easily check whether the cloud provider is disclosing his data to third
parties or whether he is using it otherwise without the user’s consent, for example.
Fig. 2 Uncertainties between cloud provider and user
S. Hauff et al.
123
These concerns have been confirmed by several interview partners on the cloud
provider side.
‘‘One aspect of virtualizing storage is that we need the trust of users. They
have to be confident that we handle their data according to their data privacy
requirements without telling them where the data is actually stored.’’
(interview partner 6)
The important role of information privacy is also supported by previous research
on e–commerce. E.g., Pavlou et al. (2007) show that purchasing behavior in online
exchange environments is influenced by information privacy concerns. Thus, we
may generalize their findings to the context of cloud computing.
Apart from information privacy concerns, interviewees also mentioned informa-
tion security concerns. Information security relates to the secure transmission and
storage of information and the prevention of information theft (Miyazaki and
Fernandez 2000). The topic is closely related to information privacy and the
uncertainties which arise are similar. Again, adverse selection and moral hazard
problems can be identified: Pre-contractually, it is difficult to examine the installed
security mechanisms of the provider and thus to judge the provider’s ability of
safeguarding all information. Post-contractually, the user cannot observe whether
the cloud provider is successfully protecting the data from hackers and espionage
attempts, for instance. This is a major concern of most users.
‘‘[…] the amount of data passed to [the provider] needs to be highly
technically secured.’’ (interview partner 4)
Also previous cloud computing opinion papers stressed the importance of data
security for cloud computing. E.g. Kaufman (2009, p. 62) emphasizes the necessity
of establishing security measures in the cloud as they ‘‘[…] offer the opportunity for
simultaneous attacks to numerous sites, and without proper security, hundreds of
sites could be comprised through a single malicious activity […]’’. These concerns
have been confirmed by several interview partners on the user and cloud provider
side.
Furthermore, availability concerns constitute another fear of users. Availability
refers to the ‘‘reliable and predictable delivery of services from the cloud’’ (Sandhu
et al. 2010). In this case, especially adverse selection problems are prevalent: The
consumer cannot evaluate a provider’s effort and investments in the reliability of his
infrastructure, e.g. whether he really mirrors all data to a second data center. Having
entered into a relationship, he can measure the actual uptime of the service.
However, he still cannot predict the future availability and thus faces uncertainty
‘‘We have in general much higher availability requirements as those offered
by typical cloud providers. […] But here is the problem: the requirements, or
that what is delivered from a cloud provider is certainly not enough. […]
However, availability and confidentiality are important criteria for choosing
our [IT] products.’’ (interview partner 3)
Exploring uncertainties in a marketplace for cloud computing
123
While availability concerns have not been connected to user’s uncertainty in
previous research, our findings indicate that, in the context of cloud computing,
availability is a major source of uncertainty (Sandhu et al. 2010).
The fourth user concern which emerged in our interviews is provider lock-in.
Since cloud products are currently not standardized, e.g. proprietary APIs are
prevalent, users may not as easily extract their data and change from one provider to
another (Armbrust et al. 2010). In this case, uncertainty arises due to adverse
selection problems: users know that choosing one cloud provider will result in a
lock-into a particular technology. Since a migration to another provider is very
costly, consumers are dependent on the provider and therefore vulnerable to price
increases and availability problems, for instance. Moreover, the triability of a
service is not given which increases the pressure on the consumer to select an
appropriate provider. These aspects are supported by several interview partners but
have also been discussed in the literature (Armbrust et al. 2010).
‘‘And then there is another important problem: how is my vendor lock-in, i.e.
how can I extract my data that belongs to me out of the data center again.’’
(interview partner 7)
To sum it up, key requirements for storage and processing power are services that
are compliant, secure, reliable and interoperable. Whether these requirements are
sufficiently met is highly dependent on the decisions of the cloud provider. As goal
conflict, fears of opportunism and asymmetric information are prevalent between
cloud provider and user, these types of uncertainties will influence the users’
decision whether to adopt and use cloud computing.
4.2 Mitigating capability of a cloud marketplace operator
E-commerce and electronic markets came with the claim of disintermediation: the
elimination of mediating roles such as wholesalers, retailers, or brokers, which
allows for a direct linkage of buyers and sellers (Wigand 1997). Currently, the cloud
landscape is organized mostly without intermediaries. Yet, as discussed in the
previous section and as outlined by Datta and Chatterjee (2008), an e-market still
faces uncertainties, even in the light of reduced search and communication costs.
This can lead to a reintermediation, the introduction of an online intermediary.
Research shows that an intermediary is seen as a source of trust and integrity of an
e-market (e.g., Giaglis et al. 2002; Son et al. 2006).
Thus, in our case, the introduction of a marketplace with a third-party
independent operator can serve as an uncertainty mitigator between users and
cloud providers. An overview of the mitigating capabilities of a cloud marketplace
is depicted in Fig. 3. Signaling can be one effective method of overcoming
uncertainties. Through the way the marketplace is implemented, it supports cloud
providers in signaling: When providers participate in the marketplace, they
automatically show that they have a certain quality because the marketplace
operator has defined admission criteria they have to fulfill; otherwise they are not
allowed to participate. Moreover, providers can just easily list all of their product
S. Hauff et al.
123
characteristics at the marketplace to show that they do not fear the comparison but
offer a sufficiently high level of quality.
‘‘I do think that this has advantages. So if a company is in such a business
case, I would say, not only information about prices, but also information
about the providers, would be a big advantage for this company. So as an
example: […] How are its computer centers distributed around the world?
Who is the provider? Where does the provider operate? What are his
priorities?’’ (interview partner 5)
Providers could still try to deceive consumers by not delivering the promised
quality. Nonetheless, at least the perceived uncertainty is reduced by increased
transparency. These two signaling methods help to reduce uncertainties. In addition,
the marketplace anticipates some of the risks and uncertainties by supporting further
signaling activities, e.g. providing warranties to consumers if services are not
available and thereby reducing the availability concerns.
‘‘So, it’s about the guarantee granted by the marketplace operator. […] There
can’t be something like, ‘No, we cannot deliver right now. You receive money,
or something like a coupon as compensation.’ […] It must be delivered in any
case; the guarantee must be given. And I think if that’s the case, then you have
again a very important added value that you sell here, namely warranty.’’
(interview partner 1)
Even more, marketing research shows that there is a strong relationship between
brands and trust. Delgado-Ballester and Munuera-Aleman (2001) confirmed that, on
the one hand, high satisfaction with a brand creates trust and, on the other hand, trust
leads to a higher commitment to and involvement with a brand. Thus, the brand
name of the marketplace operator is a good signal. MarketMaker has a good
reputation and is known as a trustworthy partner. Therefore, it does not want to risk
this image by providing a marketplace where low quality products are traded. This
is confirmed not only by the marketplace operator, it is also the perception the
Fig. 3 Uncertainty reduction through marketplace operator
Exploring uncertainties in a marketplace for cloud computing
123
interviewees from the provider and user side had. Thus, it successfully decreases
existing concerns.
‘‘So I think if the right brand operates the marketplace, simply more trust is
created. As customer, I would assume that if a provider is admitted to the
marketplace, […] the provider has been tested and meets a certain level of
quality.’’ (interview partner 10)
Monitoring and incentivizing are further methods of mitigating user concerns. On
the one hand, the marketplace operator can implement a functionality which
monitors the actual availability of a cloud service that has been traded over the
marketplace. If the availability requirements are permanently not fulfilled, they have
to face consequences like suspension. Moreover, the marketplace operator can also
monitor privacy or security breaches, at least if they become publicly known, and
exclude the involved providers. In addition, the marketplace operator could
establish feedback functionalities where users can share their experiences with
certain providers. Although this functionality is currently not planned to be
implemented on the CloudMarket, it has been mentioned in interviews as potential
uncertainty mitigation mechanism.
‘‘I think, seen from a user perspective, it would be great if I could get in touch
with other users of this cloud provider. […] You can contact another
reference, […] without the providers’ active control, which of course makes
the entire thing transparent again.’’ (interview partner 10)
While incentivizing can help to reduce all types of concerns discussed above,
monitoring can especially help to reduce information privacy and information
security concerns as well as availability concerns.
Furthermore, MarketMaker intends a standardization of IaaS-resources and a
migration functionality to easily switch from one provider to another. If this
standardization is widely accepted and the migration functionality is working, it
would have a huge impact on the user-provider relationship since it can eliminate all
lock-in concerns.
‘‘So, I believe that standardization can only be good for the market. I believe
that the entire cloud industry is currently struggling with the fact that the
forecasts are huge but volume does not turn out as planned. I think the
reluctance of the German market is right on this issue.’’ (interview partner 10)
4.3 Cloud-market-specific uncertainties
The marketplace operator also introduces new uncertainties. First, the relationship
between the marketplace operator and the consumer is analyzed: While the user still
has the same goals as before when trading resources over the marketplace, namely
buying resources based on criteria like price, privacy, security, and availability, the
marketplace operator is new in this whole picture. He is offering several services to
the user, the most important one being the matching of the consumers’ requirements
and the providers’ offerings. The marketplace operator aims at earning money with
S. Hauff et al.
123
these services with as little effort as possible. Thus, his goals are profit-oriented and
not congruent with the user’s goals. Moreover, the activities of the marketplace
operator are difficult to assess by the user, e.g. how the operator is selecting suitable
trading partners and how the monitoring of the providers’ cloud services is working.
Therefore, the chance for opportunistic behavior is given: The marketplace operator
could select providers for the matching based on where he can make the largest
profits and not based on suitability, for instance.
The uncertainties in the user-marketplace-operator relationship take many forms
and comprise adverse selection as well as moral hazard problems. Concerns about
the neutrality of the marketplace operator were issued by the interviewees.
Neutrality refers to ‘‘not taking sides, impartiality and indifference in the scientific
sense of an impartial disinterest with respect to the outcome of any conflict’’ (Hoffer
1985). In this case, the interviewees were especially concerned about the neutrality
in the selection process of the provider which cannot be assessed upfront.
‘‘I also think is important that it is a very reputable and solid marketplace
operator, one who is neutral. A marketplace operator, of which I do not expect
to make the most profit out of its services, but simply compares the companies
in a neutral way.’’ (interview partner 6)
Moreover, the interview partners mentioned concerns about the technological
ability of the marketplace operator and his ability to set standards. The marketplace
operator also confirmed the existence of these concerns. They exist especially pre-
contractually: Users cannot easily evaluate the technological ability to build the
marketplace and therefore its quality. In addition, at least shortly after the
marketplace goes live, consumers are uncertain whether the standard of the
marketplace will prevail. While the marketplace operator can reduce the availability
concerns inherent in the user-provider relationship, the users named additional
availability concerns regarding the marketplace.
‘‘One disadvantage is: what happens if I rely mostly or solely on the
marketplace? […] It is possible that the marketplace is not available and in
case a company is used to lease resources on a short-term base, then it could
run into a bottleneck.’’ (interview partner 4)
These concerns exist ex ante and ex post since the user can never determine the
future uptime of the marketplace and the operator’s efforts to ensure a high
availability rate.
Lastly, the soundness of all functionalities of the marketplace and the activities of
its operator are also seen critical because their evaluation is neither pre-
contractually nor post-contractually possible in an easy way. This also includes
concerns about information security and privacy.
‘‘[…] it must be possible to assure that the service provided by the
marketplace operator is working in terms of reliability, risk management,
market supervision and all the other topics.’’ (interview partner 12)
Concerning the cloud provider-marketplace operator relationship, the provider
still has the same goals as before: He wants to maximize his profit with minimal
Exploring uncertainties in a marketplace for cloud computing
123
efforts, while the marketplace operator wants to maximize his profits as well.
However, these goals are not congruent, since the provider wants to sell as many of
his cloud resources as possible while the marketplace wants to conduct as many
matchings as possible, yet, not necessarily with the same provider. The provider
cannot easily observe the behavior and the activities of the marketplace operator,
e.g. how the marketplace operator is selecting suitable trading partners and whether
he can really evaluate the quality of a provider and his services. Thus, the chance for
opportunistic behavior is again given.
The uncertainties which arise out of this principal-agent relationship are very
similar to the ones existent in the user-marketplace relationship. Providers are also
concerned about the neutrality of the marketplace operator, about his technological
ability, his ability to set standards as well as the soundness of the marketplace
functionalities and of the behavior of the marketplace operator.
‘‘Yes, exactly, the neutrality of the marketplace operator is of course
questionable. I do not think the operator can be neutral. […] For example,
just as an insurance broker can’t be neutral, as he also has his favorite
customers and that is just the way it is. […] With open, transparent quality
criteria, which are regularly reviewed, there will be trust.’’ (interview partner
9)
Since all these aspects have already been discussed in the context of the user-
marketplace-operator relationship, they are only mentioned here but not elaborated
in detail. In addition, the interview partners mentioned two more types of
uncertainty: The first one deals with concerns about a loss of responsibilities, e.g.
concerning the product design, the contract design, and the design of the interfaces.
This concern is existent pre- and post-contractually: The interviewees are especially
afraid of changes in the offerings’ design they are not in accord with and cannot
influence. The second uncertainty relates to a potential loss of reputation through the
participation in a malfunctioning marketplace and thus unsatisfied customers.
‘‘I would thoroughly inquire what is behind the marketplace, what kind of
intention MarketMaker has and […] elicit this significantly. It is a piece of our
reputation, which is transferred there. […] Since we have made the conscious
decision to become a serious supplier in the market […] we do not want to
mess anything up, if we go in cooperation with amateurish marketplace
operators.’’ (interview partner 7)
This concern also comprises adverse selection and moral hazard problems since
the provider can never be sure whether complications with the marketplace happen.
Figure 4 provides a comprehensive picture of all uncertainties introduced through
the marketplace operator.
4.4 Cloud-market success factors
In the following, mechanisms which can help to mitigate the uncertainty introduced
by the marketplace operator are presented. Since the mitigators are very similar for
S. Hauff et al.
123
both relationships, they are discussed in combination. A high level of transparency
constitutes the first possibility of uncertainty reduction.
‘‘To get this, it is a crucial success factor […] to make the structure of the
marketplace, our admission criteria, our SLAs and so on, transparent and also
communicate this to our all market participants.’’ (interview partner 12)
For instance, the marketplace operator can openly communicate the structure of
the marketplace, his matching mechanisms, the admission criteria, and the SLAs, to
name but a few examples. Thereby, he can show that he is a neutral instance and
that users and providers can trust in him. Moreover, the marketplace can offer a
transparent feedback mechanism where providers as well as users can see the
feedback about other participants. This enhances the perceived transparency and
neutrality. In addition, external control mechanisms of trusted third parties can
create trust in the marketplace, e.g. certificates issued by an auditing organization
with an excellent reputation.
‘‘I think everything should be a third-party-certified or it should at least be
assured that everything is above board […], in terms that nothing is tricked in
the background, for example that one provider gets more information than
another and so on. […] of course there has to be a certificate on it somehow
[…].’’ (interview partner 10)
Furthermore, the marketplace operator can offer potential participants to include
them in the conceptual as well as in the implementation phase of the marketplace as
MarketMaker currently plans to do. This constitutes a further signal: On the one
hand, all parties can represent and articulate their interests so that they might be
considered in the design. On the other hand, they can also gain detailed insights into
the marketplace design and check the soundness of the functionalities. Thus, it can
increase the trust in the marketplace. An excellent reputation is a fourth signaling
construct which can reduce uncertainty. As discussed above, an organization aims at
keeping its good reputation since it is an important asset for future business.
Fig. 4 Uncertainties between marketplace operator and participants
Exploring uncertainties in a marketplace for cloud computing
123
Moreover, especially for consumers who are currently only hesitantly using cloud
computing and for providers who have no experience with marketplaces, a
marketplace operator with a well-known brand name can establish trust in his
services.
‘‘But I think because it is MarketMaker you know that you can rely on them.
So if you can handle transactions in shares, then maybe you should be able to
do a few transactions on a cloud market.’’ (interview partner 10)
Allowing the participants to test and challenge the marketplace can also increase
confidence in the system because they do not have to believe in another party but
can test the functionalities themselves. This might not only include trying out the
functionalities but also penetration tests to assess the security of the marketplace.
‘‘Therefore, we would be happy if we could conduct security penetration tests
with the interfaces of the marketplace system, because everything is
transferred over these.’’ (interview partner 4)
This last mitigator belongs to the category of monitoring mechanisms. Figure 5
summarizes the mitigators graphically.
5 Discussion
The objective of this study is to develop a theoretical framework for understanding
uncertainties surrounding cloud markets for highly standardized computing
resources and the trade-offs they imply for an adoption decision. Overall, our
results and interpretations show that agency problems are prevalent in cloud markets
and that the principal-agent theory is a well chosen theoretical perspective to
contribute to understanding participants’ behavior. It is particularly suitable as a
theoretical scaffold for our study, since information asymmetries between cloud
provider and user play a major role in the existence of uncertainties preceding the
Fig. 5 Uncertainty mitigators between marketplace operator and participants
S. Hauff et al.
123
adoption of a cloud market by participants. The key findings that provide new
insights on uncertainties surrounding cloud computing marketplaces are discussed
in detail in the following. Subsequently, the theoretical and practical contributions
as well as suggestions for future research are presented.
5.1 Key findings
Our study corroborates previous research on cloud computing that has also
identified information privacy (e.g., Takabi et al. 2010), information security (e.g.,
Miyazaki and Fernandez 2000), availability (e.g., Sandhu et al. 2010), and lock-in
(e.g., Armbrust et al. 2010) concerns as the major sources of uncertainties in cloud
computing service scenarios. Also, the mitigating opportunities of a cloud market
operator are similar to what has been discussed in previous research on success
factors of cloud providers (e.g., Armbrust et al. 2010; Parameswaran and Chaddha
2009).
In contrast, our empirical findings provide new insights on cloud-market-specific
uncertainties that are introduced by the cloud market and insights on how these
uncertainties can be successfully managed by the marketplace operator. Based on
our findings, we are able to better assess whether a cloud market is a more attractive
alternative for cloud providers and users as opposed to a direct and bilateral cloud
provider-user business relationship. Our key findings have important implications
for the feasibility of a cloud market in a business-to-business context and the
question how organizations will obtain highly standardized IT resources in the
future.
5.1.1 The cloud market from a user perspective
Users currently face a variety of uncertainties in bilateral relationships with cloud
providers which have also been examined in previous research (Armbrust et al.
2010; Venters and Whitley 2012). Using the principal-agent perspective, we
describe in Sect. 4.1 that uncertainties exist between users and providers: Users are
especially concerned about data privacy and security, but also fear non-availability
and a provider lock-in. A marketplace operator who steps into this relationship can
help to mitigate these uncertainties through mechanisms like admission criteria and
monitoring as presented in Sect. 4.2. The acquisition of cloud resources gets easier
and more attractive for users and thus can be an improvement of the current
situation. Overall, our analysis reveals that a marketplace has a better ability to
mitigate uncertainties compared to cloud providers as the marketplace serves as a
neutral and competent instance between cloud provider and user.
Besides these positive effects, our findings also show that such an intermediary
introduces additional uncertainties from a user perspective, e.g. regarding the
functionality and availability of the marketplace or its neutrality, as discussed in
Sect. 4.3. Therefore, obtaining cloud resources becomes again more risky for users
on the marketplace. On the other hand, the market operator has abilities to reduce
these uncertainties. Our study reveals that users are more likely to adopt and use a
cloud market which operates under a trustworthy brand, they rely on well-
Exploring uncertainties in a marketplace for cloud computing
123
established external control mechanisms and favor an early integration of users in
the design and building phase of the marketplace, to name some mitigators
presented in Sect. 4.4. Overall, our study indicates that a cloud market can be an
attractive alternative for cloud users compared to a bilateral and direct business
relationship with cloud providers.
5.1.2 The cloud market from a cloud provider perspective
For cloud providers, the situation is quite different compared to the user perspective
discussed in the previous section. In a bilateral relationship with cloud users, a
provider faces only minor uncertainties like the user’s financial solvency. Thus,
being the agent in this relationship is fruitful for the provider and rather risk-free.
The introduction of an intermediary into this picture arouses many potential
uncertainties for a provider: as our analysis in Sect. 4.3 shows, providers are
especially concerned about the functionality of the marketplace as well as the
neutrality of its operator. Moreover, they fear a potential loss of responsibility and
reputation since they no longer have full control over the way and the reliability in
which their services are delivered. The uncertainty mitigators we present in Sect. 4.4
are also relevant to the cloud providers. For them, the transparency of the
marketplace mechanisms and a strong reputation are of major interest.
Overall, the cloud market causes several types of uncertainties for the provider
but offers only risk reduction for cloud users. However, some compensation or
benefit could be gained by using a cloud market. Cloud providers can use it as an
additional sales channel for customers that only rely on the CloudMarket but not on
the cloud providers’ reputation. Also, small data centers which only have limited
capacity can fulfill large orders because they have to provide only a part of the
service, the rest can be served by another provider in a way that is transparent to the
user. Thus, in the end, the provider has to decide whether these benefits really
compensate for all additionally introduced uncertainties and whether he wants to
participate in the marketplace. Therefore, seeing it from an uncertainty perspective,
the marketplace presents a less attractive alternative to the current bilateral
relationships.
5.2 Theoretical contribution and implications for research
Our paper presents the first empirical study on a conceptualized multi-vendor,
multi-buyer, third party operated marketplace for cloud computing. Some concep-
tual studies have been carried out on the technical feasibility of such markets (e.g.,
Buyya et al. 2008; Schnizler et al. 2008). However, the conceptual evaluation of
such market mechanisms showed that the clearing of the market is far too complex
to carry it out in practice in the way the authors propose. Also, major issues of data
privacy, risk mitigation and service level enforcement remain open. Our study
shows that the motives for adopting a transparent, two-sided and independently
managed marketplace for cloud computing are distinct to those identified in
previous research on e-market adoption and that uncertainties between cloud
provider, user and market operator play a major role for the feasibility of such a
S. Hauff et al.
123
market place. Even more, while the marketplace is an attractive alternative to the
bilateral relationships for users, providers also experience new uncertainties in these
constellations. Nonetheless, if satisfying compensation can be offered to providers,
a cloud marketplace might be a promising way of trading cloud computing
resources in the future. Yet, this trilateral contract as efficient governance structure
might contradict some elements of the transaction cost theory: The theory says that
trilateral contracts are only the most suitable solution in case of occasional
transactions and mixed or idiosyncratic asset specificity (Williamson 1975). The
marketplace, however, is also used for frequent transactions and the related asset
specificity is non-specific or mixed due to the intended high degree of standard-
ization on the marketplace. Thus, this tentative finding should be subject of further
investigations in the near future.
Overall, we provide three major-contributions to the literature on cloud
computing. First, our study shows that agency problems are prevalent on cloud
markets and that the principal-agent theory is a fruitful perspective to study these
uncertainties. Since the perspective has hardly been used to study the adoption of
e-markets, we encourage future research to re-examine other theoretical e-market
contexts based on this perspective. Second, while cloud markets have capabilities to
mitigate uncertainties between cloud provider and user, they also cause new
uncertainties in this relationship. This supplements previous studies which have
emphasized and theorized the mitigating capabilities of e-markets but did not focus
on their negative effects on uncertainty perceptions of participants. Third and as
presented in the previous section, our study presents a second-level theoretical
understanding on uncertainties surrounding cloud markets and thereby sheds light
on the trade-offs involved in the decision to adopt a cloud market by cloud providers
and users. The framework can be utilized by future research to study the role of
uncertainty for other relevant behavioral outcomes in the context of cloud
computing and cloud markets in particular.
5.3 Implications for practice
Several implications for practitioners can be given. First, the framework can help
market operators in the design of the marketplace. Since uncertainty demonstrably
plays a dominant role, we want to raise market operators’ awareness for this topic:
They can reduce the uncertainties inherent in the bilateral relationships between
users and providers like offering warranties and monitoring. Thus, it is advanta-
geous to clearly communicate all mitigating capabilities a marketplace offers.
Moreover, the marketplace’s potential of uncertainty reduction is also an interesting
aspect for cloud providers since they are not only competing for current cloud users
but also for new user segments which are currently not using cloud services due to
the existing risks.
On the other hand, market operators have to keep in mind that they and their
marketplaces are also a source of uncertainty. Therefore, they should concentrate on
mitigating the concerns, e.g. by building up trust and openly and transparently
communicating all functionalities of a marketplace. Moreover, an early integration
of potential participants in the conceptual phase can increase the probability of
Exploring uncertainties in a marketplace for cloud computing
123
acceptance. While the overall degree of uncertainty is reduced for users, market
operators should carefully consider the compensations they offer to providers since
they face a higher degree of uncertainty when participating in the marketplace
compared to trading in bilateral relationships.
The third implication deals with the differences of cloud users and cloud
providers. Market operators should address both groups separately as they have
different concerns when participating in the marketplace. Therefore, market
operators should carefully think about how to present the idea of their marketplace
to each of these groups in order to get the best out of each meeting or advertising
campaign, for instance.
References
AIS (2011) Association of information systems, Georgia. http://start.aisnet.org/?SeniorScholarBasket.
Accessed 30 Jan 2013
Akerlof GA (1970) The market for ‘lemons’: quality uncertainty and the market mechanism. Q J Econ
84:488–500. doi:10.2307/1879431
Alford T, Morton G (2009) The economics of the cloud—addressing the benefits of infrastructure in the
cloud. http://www.boozallen.com/media/file/Economics-of-Cloud-Computing.pdf. Accessed 27 Nov
2013
Angeles R (2000) Revisiting the role of internet-EDI in the current electronic commerce scene. Logist Inf
Manag 13:45–57. doi:10.1108/0957605001030639
Appelbaum PS, Gold A (2010) Psychiatrists’ relationships with industry: the principal-agent problem.
Harv Rev Psychiatry 18:255–265. doi:10.3109/10673229.2010.507038
Armbrust M, Fox A, Griffith R, Joseph AD, Katz R, Konwinski A, Lee G, Patterson D, Rabkin A, Stoica
I, Zaharia M (2010) A view of cloud computing. Commun Assoc Comput Mach 53:50–58. doi:10.
1145/1721654.1721672
Bakos JY (1991) A strategic analysis of electronic marketplaces. Manag Inf Syst Q 15:295–310. doi:10.
2307/249641
Bakos JY (1997) Reducing buyer search costs: implications for electronic marketplaces. Manag Sci
43:1676–1692. doi:10.1287/mnsc.43.12.1676
Bakos JY (1998) The emerging role of electronic marketplaces on the internet. Commun Assoc Comput
Mach 41:35–42. doi:10.1145/280324.280330
BITKOM (2012) Bundesverband Informationswirtschaft, Telekommunikation und neue Medien e.V.,
Berlin. http://www.bitkom.org/de/markt_statistik/64086_71376.aspx. Accessed 30 Jan 2013
Breskovic I, Altmann J, Brandic I (2012) Creating standardized products for electronic markets. Future
Gener Comput Syst 29:1000–1011. doi:10.1016/j.future.2012.06.007
Buyya R, Yeo CS, Venugopal S (2008) Market-oriented cloud computing: vision, hype, and reality for
delivering IT services as computing utilities. Paper presented at the 10th IEEE international
conference on high performance computing and communications, Dalian, China, 25–27 Sept 2008
Buyya R, Pandey S, Vecchiola C (2009a) Cloudbus toolkit for market-oriented cloud computing. In:
Jaatun MG, Zhao G, Rong C (eds) Cloud computing. Springer, Berlin, Heidelberg, pp 24–44
Buyya R, Yeo CS, Venugopal S, Broberg J, Brandic I (2009b) Cloud computing and emerging IT
platforms: vision, hype, and reality for delivering computing as the 5th utility. Future Gener Comput
Syst 25(6):599–616. doi:10.1016/j.future.2008.12.001
Calheiros RN, Toosi AN, Vecchiola C, Buyya R (2012) A coordinator for scaling elastic applications
across multiple clouds. Future Gener Comput Syst 28:1350–1362. doi:10.1016/j.future.2012.03.010
Carr NG (2003) IT doesn’t matter. Harv Bus Rev 81:1–32. doi:10.1109/EMR.2004.25006
Chang HH, Wong KH (2010) Adoption of e-procurement and participation of e-marketplace on firm
performance: trust as a moderator. Inf Manag 47:262–270. doi:10.1016/j.im.2010.05.002
Coase RH (1937) The Nature of the Firm. Economica 4:386–405. doi:10.1111/j.1468-0335.1937.tb00002.x
S. Hauff et al.
123
Datta P, Chatterjee S (2008) The economics and psychology of consumer trust in intermediaries in
electronic markets: the EM trust framework. Eur J Inf Syst 17:12–28. doi:10.1057/palgrave.ejis.
3000729
Delgado-Ballester E, Munuera-Aleman JL (2001) Brand trust in the context of consumer loyalty. Eur J
Mark 35:1238–1258. doi:10.1108/EUM0000000006475
Dimoka A, Hong Y, Pavlou P (2012) On product uncertainty in online markets: theory and evidence.
Manag Inf Syst Q 36:395–426
Eisenhardt KM (1989a) Agency theory: an assessment and review. Acad Manag Rev 14:57–72. doi:10.
2307/258191
Eisenhardt KM (1989b) Building theories from case study research. Acad Manag Rev 14:532–550.
doi:10.5465/AMR.1989.4308385
Foster I, Zhao Y, Raicu I, Lu S (2008) Cloud computing and grid computing 360-degree compared. Paper
presented at grid computing environments workshop, 2008. GCE’08, Austin, US, 12–16 Nov 2008
Garg SK, Vecchiola C, Buyya R (2011) Mandi: a market exchange for trading utility and cloud
computing services. J Supercomput 64:1153–1174. doi:10.1007/s11227-011-0568-6
Gartner (2012) Gartner says worldwide cloud services market to surpass $109 billion in 2012. http://
www.gartner.com/newsroom/id/2163616. Accessed 30 Jan 2013
Gengatharen DE, Standing C (2005) A framework to assess the factors affecting success or failure of the
implementation of government-supported regional e-marketplaces for SMEs. Eur J Inf Syst
14:417–433. doi:10.1057/palgrave.ejis.3000551
Giaglis GM, Klein S, O’Keefe RM (2002) The role of intermediaries in electronic marketplaces:
developing a contingency model. Inf Syst J 12:231–246. doi:10.1046/j.1365-2575.2002.00123.x
Gordon RL (1980) Interviewing, strategy, techniques and tactics. Dorsey Press, Homewood, IL
Grance T, Mell PM (2011) The NIST definition of cloud computing. US Department of Commerce. http://
www.nist.gov/manuscript-publication-search.cfm?pub_id=909616. Accessed 29 Feb 2012
Grilo A, Jardim-Goncalves R (2013) Cloud-marketplaces: distributed e-procurement for the AEC sector.
Adv Eng Inf 27:160–172. doi:10.1016/j.aei.2012.10.004
Hoffer A (1985) Toward a definition of psychoanalytic neutrality. J Am Psychoanal Assoc 33:771–795.
doi:10.1177/000306518503300402
Holton JA (2007) The coding process and its challenge. In: Charmaz K, Bryant A (eds) The sage
handbook of grounded theory. Sage Publications, London, pp 265–290
Jensen MC, Meckling WH (1976) Theory of the firm: managerial behavior, agency costs and ownership
structure. J Financ Econ 3:305–360. doi:10.1016/0304-405x(76)90026-x
Joo YB, Kim YG (2004) Determinants of corporate adoption of e-marketplace: an innovation theory
perspective. J Purch Supply Manag 10:89–101. doi:10.1016/j.pursup.2004.01.001
Kaufman LM (2009) Data security in the world of cloud computing. IEEE Secur Priv 7:61–64. doi:10.
1109/MSP.2009.87
Krieger O, McGachey P, Kanevsky A (2010) Enabling a marketplace of clouds: VMware’s vCloud
Director. SIGOPS Oper Syst Rev 44:103–114. doi:10.1145/1899928.1899942
Lee A (1991) Integrating positivist and interpretive approaches to organizational research. Organ Sci
2:342–365. doi:10.1287/orsc.2.4.342
Lenk A, Klems M, Nimis J, Tai S, Sandholm T (2009) What’s inside the cloud? An architectural map of
the cloud landscape. Paper presented at proceedings of the 2009 ICSE workshop on software
engineering challenges of cloud computing, Washington, USA, 23 May 2009
Li H, Jeng JJ (2010) CCMarketplace: a marketplace model for a hybrid cloud. Paper presented at
proceedings of the 2010 conference of the center for advanced studies on collaborative research,
Toronto, Canada, 1–4 Nov 2010
Malkiel BG, Fama EF (1970) Efficient capital markets: a review of theory and empirical work. J Financ
25:383–417. doi:10.1111/j.1540-6261.1970.tb00518.x
Malone TW, Yates J, Benjamin RI (1987) Electronic markets and electronic hierarchies. Commun Assoc
Comput Mach 30:484–497. doi:10.1145/214762.214766
Mishra DP, Heide JB, Cort SG (1998) Information asymmetry and levels of agency relationships. J Mark
Res 35:277–295. doi:10.2307/3152028
Miyazaki AD, Fernandez A (2000) Internet privacy and security: an examination of online retailer
disclosures. J Public Policy Mark 19:54–61. doi:10.1509/jppm.19.1.54.16942
Mugge D (2011) The European presence in global financial governance: a principal-agent perspective.
J Eur Public Policy 18:383–402. doi:10.1080/13501763.2011.551075
Nelson P (1970) Information and consumer behavior. J Polit Econ 78:2311–2329
Exploring uncertainties in a marketplace for cloud computing
123
Parameswaran AV, Chaddha A (2009) Cloud interoperability and standardization. SETLabs Brief
7:19–26
Patton MQ (2002) Qualitative research and evaluation methods, 3rd edn. Sage, Thousand Oaks
Pavlou PA, Liang H, Xue Y (2007) Understanding and mitigating uncertainty in online exchange
relationships: a principal-agent perspective. Manag Inf Syst Q 31:105–136
Quaddus M, Hofmeyer G (2007) An Investigation into the factors influencing the adoption of B2B trading
exchanges in small businesses. Eur J Inf Syst 16:202–215. doi:10.1057/palgrave.ejis.3000671
Rao A, Ruekert RW (1994) Brand alliances as signals of product quality. Sloan Manag Rev 36:87–97
Rask M, Kragh H (2004) Motives for e-marketplace participation: differences and similarities between
buyers and suppliers. Electron Mark 14:270–283. doi:10.1080/10196780412331311720
Rossbach C, Welz B (2011) Survival of the fittest—how europe can assume a leading role in the cloud.
Roland Berger strategy consultans. http://www.rolandberger.com/media/pdf/Roland_Berger_
Cloud_Ecosystem_D_20111130.pdf. Accessed 19 June 2012
Sandhu R, Boppana R, Krishnan R, Reich J, Wolff T, Zachry J (2010) Towards a discipline of mission-
aware cloud computing. In Proceedings of the 2010 ACM workshop on cloud computing security
workshop, New York, US. doi:10.1145/1866835.1866839
Sarker S, Sarker S (2009) Exploring agility in distributed information systems development teams: an
interpretive study in an offshoring context. Inf Syst Res 20:440–461. doi:10.1287/isre1090.0241
Sarker S, Sahaym A, Bjorn-Andersen N (2012) Exploring value cocreation in relationships between an
ERP Vendor and its partners: a revelatory case study. Manag Inf Syst Q 36:317–338
Schepers J, Falk T, de Ruyter K, de Jong AD, Hammerschmidt M (2012) Principles and principals: do
customer stewardship and agency control compete or complement when shaping frontline employee
behavior? J Mark 76:1–20. doi:10.1509/jm.11.0112
Schnizler B, Neumann D, Veit D, Weinhardt C (2008) Trading grid services—a multi-attribute
combinatorial approach. Eur J Oper Res 187:943–961
Smith A (1937[1776]) The Wealth of Nations. In: Cannan E (ed) Reprint, 1904. Modern Library, New
York
Son JY, Tu L, Benbasat I (2006) A descriptive content analysis of trust-building measures in B2B
electronic marketplaces. Commun Assoc Inf Syst 18:99–128
Stockdale R, Standing C (2004) Benefits and barriers of electronic marketplace participation: an SME
perspective. J Enterp Inf Manag 17:301–311. doi:10.1108/17410390410548715
Strausz R (1997) Delegation of monitoring in a principal-agent relationship. Rev Econ Stud 64:337–357.
doi:10.2307/2971717
Takabi H, Joshi JBD, Ahn GJ (2010) Security and privacy challenges in cloud computing environments.
IEEE Secur Priv 8:24–31. doi:10.1109/MSP.2010.186
Tao YH, Chen CP, Chang CR (2007) Unmet adoption expectation as the key to e-marketplace failure: a
case of Taiwan’s steel industry. Ind Mark Manag 36:1057–1067. doi:10.1016/j.indmarman.2006.08.
004
Venters W, Whitley EA (2012) A critical review of cloud computing: researching desires and realities.
J Inf Technol 27:179–197. doi:10.1057/jit.2012.17
Vermillion LJ, Lassar WM, Winsor RD (2002) The Hunt-Vitell general theory of marketing ethics: can it
enhance our understanding of principal-agent relationships in channels of distribution? J Bus Ethics
41:267–285. doi:10.1023/A:1021284922440
Walsham G (1995) Interpretive case studies in IS research: nature and method. Eur J Inf Syst 4:74–81.
doi:10.1057/ejis
Westin AF (1967) Privacy and freedom. Atheneum Press, New York
Wigand RT (1997) Electronic commerce: definition, theory, and context. Inf Soc 13:1–16. doi:10.1080/
019722497129241
Williamson OE (1975) Markets and hierarchies. The Free Press, New York
Williamson OE (1985) The economic institutions of capitalism: firms, markets, relational contracting.
The Free Press, New York
Yin RK (2009) Case study research: design and methods, 4th edn. SAGE Publications, Thousand Oaks
Zhang Q, Cheng L, Boutaba R (2010) Cloud computing: state-of-the-art and research challenges.
J Internet Serv Appl 1:7–18. doi:10.1007/s13174-010-0007-6
S. Hauff et al.
123